9.1 The stratigraphy and chronology of the Pleistocene sediments at Warsash
The terrace stratigraphy of the River Test at Warsash has been reassessed as described above, with revised attributions to the Hamble, Mottisfont/Lower Warsash, Belbin/Upper Warsash and Ganger Wood/Mallards Farm Terraces in the area (Figures 8 and 11). The new stratigraphic detail has provided a more robust framework for the spatial and temporal distribution of the Palaeolithic record. These changes are significant for understanding the characteristics of the archaeology of the Warsash region as discussed in section 9.2. The revisions have also allowed a broader reassessment of the Palaeolithic archaeology of Warsash and its place in the Lower- Middle Pleistocene settlement history of southern Britain (Davis et al. 2016).
Two of the luminescence ages produced here are consistent with previous age determinations and stratigraphy. Samples HAP10-03Qz (Mottisfont/Lower Warsash Terrace) and BRW08-02Qz (Hamble Terrace) are stratigraphically consistent, although uncertainties overlap. BRW08-02Qz is comparable to the youngest age already reported for the Hamble Terrace in the Test region of 203 ±17.7 ka (MIS 7c- 6) (Bates et al. 2004), although one of the Mottisfont/Lower Warsash Terrace age calculations (HAP10-03Qz) also falls within the PASHCC study’s range of MIS 8-6 for the Hamble Terrace (with a weighted mean of 217 ±22 ka (MIS 7)). BRW08- 02Qz and the PASHCC results indicate a MIS 7 age for the deposition of the Hamble Terrace. The age estimate for HAP10-03Qz is comparable to the attribution of MIS 9- 8 for the Mottisfont terrace at Dunbridge, based on OSL-dates (Harding et al. 2012). The Mottisfont Terrace appears correlative to the Lower Warsash Terrace based on luminescence dating and the long profile presented above (Figure 11). The results produced here suggest aggradation of the Mottisfont/Lower Warsash Terrace during MIS 8 followed by the Hamble Terrace during MIS 7. An age estimate of MIS 8 for the aggradation of the Mottisfont/Lower Warsash Terrace may suggest that the Warsash handaxe assemblage derives from MIS 9 or earlier. It is noted that further dating studies are required to address the methodological issues identified above; therefore, any correlations attempted here are tentative and have to be treated with caution.
Finally, the quartz sample HAP10-02Qz and feldspar sample WAC10-03Fs both appear too young. Both derive from the same fluvial terrace as HAP10-03Qz, located in neighbouring gravel pits at Warsash at similar altitude, yet HAP10-02Qz produced an indicative age of 119 ±10.7 ka (MIS 5e-5d) and WAC10-03Fs an age of 55 ±5.4 ka
(MIS 4-3). This may be due to inhomogeneity of samples not detected by ICP-MS analysis or other problems that are inherent in the bedrock geology of the catchment. The dose rates calculated for HAP10-02Qz and WAC10-03Fs are higher than those calculated in the PASCHH studies (Schwenninger et al. 2006, 2007), which reported rates of 0.81-1.19 (Gy ka-1) for the majority (10) of Test samples; they are also around 3 to 4 times those of BRW08-02Qz and HAP10-03Qz (Table 7). The PASCHH project did produce two samples with higher rates of 1.61 and 2.31 (Gy ka-1), comparable to those for samples HAP10-02Qz and WAC10-03Fs, but these were based on Neutron Activation Analysis (NAA) rather than in situ gamma spectrometry. For sample HAP10-02Qz, applying a dose rate towards the lower range measured by PASHCC (~0.81-0.90 Gy ka-1) would produce an age estimate similar to HAP10- 03Qz of around 230-250 ka. For sample WAC10-03Fs a lower dose rate of ~0.45- 0.50 Gy ka-1 would be required.
Each of the three recent attempts to date terraces in the Test sequence, PASHCC, Harding et al. (2012) and this study, have encountered issues with the results obtained. This study acquired likely minimum ages of 229 ±23.7 (MIS 8-7) for the Mottisfont/Lower Warsash Terrace and 200 ±22.8 (MIS 7-6) for the Hamble Terrace, comparable with results from the PASHCC project. Two rejected ages appear to have unrealistically high dose rates. The PASHCC ages are acknowledged to be problematic above the lowest terraces sampled (Bates and Briant 2009; Briant et al. 2012), In calculating the slightly later attribution of MIS 9-8 for the Mottisfont/Lower Warsash Terrace at Dunbridge, Harding et al. (2012) excluded four ages ranging from 456 ±101 ka to 393 ±62 ka as being unreliable. The remaining four ages ranged from 335 ±45 ka to 262 ±43 ka (MIS 11-7) and were not stratigraphically consistent. The methods applied here indicate a high rejection rate of samples and aliquots within samples which would not have been detected in the PASHCC and Harding et al. (2012) studies. Further work is needed to investigate why so many problems have been encountered in attempts to use luminescence dating in the Test Valley.
Understanding the chronology of the Test sequence remains problematic above the Mottisfont/Lower Warsash Terrace. Previous work by Bates et al. (2004; cf. Westaway et al. 2006) has proposed a correlation between the Nursling Terrace of the River Test and a cold-stage before or after the MIS 13 Goodwood/Slindon Raised Beach (Roberts and Parfitt 1999) (i.e. MIS 14 or 12). In such a scenario it is likely that at least the Bitterne Terrace and above of the Test sequence were deposited prior to MIS 13. More chronological tie-points above the Mottisfont/Lower Warsash Terrace are required to construct a robust stratigraphic sequence for the Test.
9.2 Implications for the terrace stratigraphy and Palaeolithic archaeology of the River Test
The terrace stratigraphy of the River Test has been reassessed as described above, with revised correlations of terrace levels between BGS sheets 299 (Winchester) and 315 (Southampton) (Figure 11). The results suggest agreement with the correlation of Lower and Upper Terraces at Warsash with the Mottisfont and Belbin Terraces upstream as per Harding et al. (2012) (cf. the PASHCC model). The Hamble and Ganger Wood/Mallards Moor Terraces downstream are not recognised in the Dunbridge area. Correlations have also been proposed for the Nursling, Bitterne and
Midanbury/Rownham’s Farm Terraces while recognising that the latter two terraces are poorly represented in the dataset.
Table 10 shows the revision of terrace attribution for some significant Palaeolithic archaeological sites located at Warsash, Dunbridge and elsewhere in the Test region. The revisions proposed by this study have a number of implications for the understanding and interpretation of the archaeological record. Two implications in particularly are of consequence: the relationship of terraces of the River Test upstream at Dunbridge and downstream at Warsash; and the terrace attributions of individual assemblages in the Warsash area.
The earliest archaeological evidence in the Test region, and potentially the Solent region as a whole, is the three handaxes found at Towns Pit, Southampton Common (Davis 2015; Table 10), which retains its attribution to the Midanbury/Rownham’s Farm Terrace here. The upstream correlation of River Test terraces between the Southampton BGS map sheet and the Winchester sheet favoured here results in a reattribution of the Great Copse, Mottisfont artefacts (Table 10) from the Bitterne Terrace (Terrace 7 of PASHCC) to the Nursling Terrace. Two major sites at Romsey remain in the Belbin/Upper Warsash Terrace in the revised terrace scheme (Table 10). The important sites at Dunbridge and Kimbridge are attributed to Belbin/Upper Warsash and Mottisfont/Lower Warsash respectively (Table 10), and downstream sites at Warsash remain in the Mottisfont/Lower Warsash and Hamble Terraces (Table 10) as in previous schemes (Edwards and Freshney 1987; Westaway et al. 2006).
The revisions to the terrace mapping in and around Warsash enable some of the Warsash archaeological material to be assigned to specific terraces. As discussed previously, the majority of the Warsash record lacks locality data, with just a small amount that has a specific pit recorded. Davis’s (2013; Davis et al. 2016) recent review has established that the four gravel pits discussed by Burkitt et al. (1939) are all located in areas of the Mottisfont/Lower Warsash Terrace (Figure 2). Therefore, all of the Mogridge Collection that can be demonstrated to have been collected prior to 1939 can be assigned to the Mottisfont/Lower Warsash Terrace. Further, all gravel pits in the Warsash area prior to 1945 were restricted to areas of the Mottisfont/Lower Warsash Terrace (Davis et al. 2016). So any artefacts collected prior to 1945 can be assigned to Mottisfont/Lower Warsash. After 1945, quarrying in the region exploited gravels of the Hamble and Mottisfont/Lower Warsash Terraces. Therefore artefacts with only a general Warsash provenance recovered after 1945 cannot be assigned to a specific terrace. On this basis, 254 handaxes and 30 Levallois artefacts can be associated with the Mottisfont/Lower Warsash Terrace, 51 handaxes with the Hamble Terrace, while 194 handaxes and 4 Levallois artefacts cannot be assigned to a specific terrace.
It is therefore likely that the Levallois material from Warsash is exclusively associated with the Mottisfont/Lower Warsash Terrace. It is also clear from the condition of the artefacts – the majority of the Levallois material is fresh and patinated, contrasting the typically rolled and stained handaxes – that the Levallois assemblage has a different taphonomic history to the handaxes associated with the same terrace (Ashton & Hosfield 2010; Davis et al. 2016). The high degree of rolling and staining among the Mottisfont/Lower Warsash Terrace handaxes strongly suggests that they originated
within terrace gravels, an assertion that is supported by the observations of Burkitt et al. (1939), who stated that two of their three series of handaxes were recovered from the basal gravels. If an MIS 8 age for the Mottisfont/Lower Warsash Terrace is accepted, these are likely to have been reworked from earlier deposits of at least MIS 9 age.
With regards to the Levallois material, Burkitt et al. suggest that at least some of it originated in fine-grained deposits overlying the terrace gravels and therefore post- dates terrace formation. The fresh condition of the artefacts fits with this interpretation. A similar situation is found at several sites of the Middle Thames, such as Creffield Road and Yiewsley (Scott et al. 2011). There, fresh Levallois artefacts have been observed to rest on, or in sediments that overlie, Lynch Hill gravels that contain rolled handaxes (Brown 1889, 1895). Ashton et al. (2003) argue that the Levallois material was either discarded on the margins of the floodplain prior to downcutting during late MIS 8, or discarded post-downcutting on the terrace surface adjacent to the new floodplain during MIS 7. If a parallel situation is found at Warsash, then the fresh Levallois material may date to late MIS 8 or MIS 7.
Table 10. Major Palaeolithic artefact site locations as assigned in previous schemes and the revised terrace stratigraphy of the River Test. Site location precision key: [A] Accurate; [E] Estimated; [G] General. Artefact numbers key: H Handaxes; L Levallois; O Other. Previous terrace scheme and previous MIS model key: 1 Edwards & Freshney 1987); 2 Westaway et al. (2006); 3 PASHCC (Bates et al. 2004, 2007; Bates and Briant 2009); 4 Harding et al. (2012). Westaway et al. (2006)/ Harding et al. (2012) terrace nomenclature: Mottisfont/LW: Mottisfont/Lower Warsash; Belbin/UW: Belbin/Upper Warsash. Attributions in bold indicate revised terrace correlations and/or MIS age modelling as discussed in the text. Site location and artefact data from Davis (2013). Site location [Precision] Artefacts H L O Previous terrace schemes Harding et al. MIS model Revised terrace scheme Probable MI Stage [Range] Town Pits, [A]
Southampton Common 3 0 0 Terrace 8 1 Rownham’s Farm 4 14 4 Midanbury/ Rownham’s Farm ?16-15 [>13] Great Copse, Mottisfont [A] 1 0 3 Terrace 7 3 Not specified 4 - Nursling ?14-12
Chivers Gravel Pit, Romsey Extra [A]
100 3 18 Terrace 4 1 Belbin/UW 2, 4 9b 4 Belbin/ Upper Warsash ?9 [12-9] Belbin's Pit, Romsey
Extra [A] 200 3 9 Terrace 4 1 Belbin/UW 2, 4 9b 4 Belbin/ Upper Warsash ?9 [12-9] Dunbridge:
Dunbridge Hill [A] 1000 5 0
Belbin/UW 2, 4
Terrace 5 3 9b 4
Belbin/
Upper Warsash ?9 [12-9] Hatt Hill [E] 1 0 0
RMC Gravel Pit [A] 0 0 5 Kimbridge, Mottisfont [A] 77 0 9 Mottisfont/LW 2, 4 Terrace 4 3 8 4 Mottisfont/ Lower Warsash 8 [8-7] Warsash:
Fleet End Pit [A] 20 13 2
Terrace 3 1, 3
Mottisfont/LW 4 8 4
Mottisfont/
Lower Warsash 8 [8-7] New Pit [A] 15 4 0
Park’s Pit [A] 10 0 0 Button’s Pit [E] 0 0 1 Dyke’s Pit [A] 2 0 0 Hook Lane [G] 1 0 0
Newbury’s Pit [A] 6 0 1 Terrace 2 1, 3 Hamble 2, 4 6 4 Mottisfont/ Lower Warsash 8 [8-7] Warsash: General 200 13 43 Terrace 2 or 3 1, 3 Hamble or Mottisfont/ LW 4 6 or 8 4 Mottisfont/ Lower Warsash 8 [8-7]
Site location [Precision] Artefacts H L O Previous terrace schemes Harding et al. MIS model Revised terrace scheme Probable MI Stage [Range] Warsash: General 194 4 34 Terrace 2 or 3 1, 3 Hamble or Mottisfont/ LW 4 6 or 8 4 ? Hamble or Mottisfont/ Lower Warsash 7 [7-6] or 8 [8-7]
9.3 Methodological approaches to constructing long profile projections and correlations
In regions where diagnostic lithological, biostratigraphical or chronological data are scarce, whether due to minimal variations in clast input into the fluvial system over time, preservation issues, or the availability of sedimentary exposures or datasets, terrace remnants may be correlated by means of altitudinal position along the river’s palaeo-course alone (Briant et al. 2012). Such long profile correlations of terrace bodies are usually based on downstream projections of approximately straight or slightly concave upward gradients (Gibbard 1985; Briant et al. 2012). This has been the case in the Test Valley, where interpretation of the terrace stratigraphy and important downstream correlations of often fragmentary terrace units has been reliant on limited, and methodologically different, datasets as discussed above. Two recent terrace stratigraphies have been constructed for the River Test using contrasting data to describe the terrace deposits. Post-depositional modification may affect methods based on modern terrace ‘surfaces’ (i.e. ground level), which may not be representative of former terrace aggradations. Methods based on the thickness of underlying sedimentary deposits need to account for topographical variation in the palaeo-floodplain or changing terrace thickness between the front and back of an outcrop. Where datasets are sufficiently large, an assessment can be made on the representative nature of each sedimentary record in relation to the framework as a whole. Comparison of Figures 10 and 11 shows that a more robust terrace stratigraphy can be constructed by use of sedimentary data (in this case bedrock elevation and terrace deposit thickness) rather than ground surface data. Such an approach is dependent on sufficient data-coverage and it has been demonstrated that the use of GPR can be an effective method to close larger data gaps. The method is time efficient and allows extensive data capture. Synthetic boreholes (Hatch 2014) can be used to summarise linear datasets and enable integration with other data types, such as borehole records and sedimentary logs.